2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2021 Ng Peng Nam Sean
5 * Copyright (c) 2022 Alexander V. Chernikov <melifaro@FreeBSD.org>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * This file contains socket and protocol bindings for netlink.
33 #include <sys/param.h>
34 #include <sys/kernel.h>
35 #include <sys/malloc.h>
37 #include <sys/rmlock.h>
38 #include <sys/domain.h>
41 #include <sys/protosw.h>
44 #include <sys/socket.h>
45 #include <sys/socketvar.h>
46 #include <sys/sysent.h>
47 #include <sys/syslog.h>
48 #include <sys/priv.h> /* priv_check */
50 #include <netlink/netlink.h>
51 #include <netlink/netlink_ctl.h>
52 #include <netlink/netlink_var.h>
54 #define DEBUG_MOD_NAME nl_domain
55 #define DEBUG_MAX_LEVEL LOG_DEBUG3
56 #include <netlink/netlink_debug.h>
57 _DECLARE_DEBUG(LOG_DEBUG);
59 _Static_assert((NLP_MAX_GROUPS % 64) == 0,
60 "NLP_MAX_GROUPS has to be multiple of 64");
61 _Static_assert(NLP_MAX_GROUPS >= 64,
62 "NLP_MAX_GROUPS has to be at least 64");
64 #define NLCTL_TRACKER struct rm_priotracker nl_tracker
65 #define NLCTL_RLOCK(_ctl) rm_rlock(&((_ctl)->ctl_lock), &nl_tracker)
66 #define NLCTL_RUNLOCK(_ctl) rm_runlock(&((_ctl)->ctl_lock), &nl_tracker)
68 #define NLCTL_WLOCK(_ctl) rm_wlock(&((_ctl)->ctl_lock))
69 #define NLCTL_WUNLOCK(_ctl) rm_wunlock(&((_ctl)->ctl_lock))
71 static u_long nl_sendspace = NLSNDQ;
72 SYSCTL_ULONG(_net_netlink, OID_AUTO, sendspace, CTLFLAG_RW, &nl_sendspace, 0,
73 "Default netlink socket send space");
75 static u_long nl_recvspace = NLSNDQ;
76 SYSCTL_ULONG(_net_netlink, OID_AUTO, recvspace, CTLFLAG_RW, &nl_recvspace, 0,
77 "Default netlink socket receive space");
79 extern u_long sb_max_adj;
80 static u_long nl_maxsockbuf = 512 * 1024 * 1024; /* 512M, XXX: init based on physmem */
81 static int sysctl_handle_nl_maxsockbuf(SYSCTL_HANDLER_ARGS);
82 SYSCTL_OID(_net_netlink, OID_AUTO, nl_maxsockbuf,
83 CTLTYPE_ULONG | CTLFLAG_RW | CTLFLAG_MPSAFE, &nl_maxsockbuf, 0,
84 sysctl_handle_nl_maxsockbuf, "LU",
85 "Maximum Netlink socket buffer size");
88 * Looks up a nlpcb struct based on the @portid. Need to claim nlsock_mtx.
89 * Returns nlpcb pointer if present else NULL
92 nl_port_lookup(uint32_t port_id)
96 CK_LIST_FOREACH(nlp, &V_nl_ctl->ctl_port_head, nl_port_next) {
97 if (nlp->nl_port == port_id)
104 nl_add_group_locked(struct nlpcb *nlp, unsigned int group_id)
106 MPASS(group_id <= NLP_MAX_GROUPS);
109 /* TODO: add family handler callback */
110 if (!nlp_unconstrained_vnet(nlp))
113 nlp->nl_groups[group_id / 64] |= (uint64_t)1 << (group_id % 64);
117 nl_del_group_locked(struct nlpcb *nlp, unsigned int group_id)
119 MPASS(group_id <= NLP_MAX_GROUPS);
122 nlp->nl_groups[group_id / 64] &= ~((uint64_t)1 << (group_id % 64));
126 nl_isset_group_locked(struct nlpcb *nlp, unsigned int group_id)
128 MPASS(group_id <= NLP_MAX_GROUPS);
131 return (nlp->nl_groups[group_id / 64] & ((uint64_t)1 << (group_id % 64)));
135 nl_get_groups_compat(struct nlpcb *nlp)
137 uint32_t groups_mask = 0;
139 for (int i = 0; i < 32; i++) {
140 if (nl_isset_group_locked(nlp, i + 1))
141 groups_mask |= (1 << i);
144 return (groups_mask);
148 * Broadcasts message @m to the protocol @proto group specified by @group_id
151 nl_send_group(struct mbuf *m, int num_messages, int proto, int group_id)
153 struct nlpcb *nlp_last = NULL;
157 IF_DEBUG_LEVEL(LOG_DEBUG2) {
158 struct nlmsghdr *hdr = mtod(m, struct nlmsghdr *);
159 NL_LOG(LOG_DEBUG2, "MCAST mbuf len %u msg type %d len %u to group %d/%d",
160 m->m_len, hdr->nlmsg_type, hdr->nlmsg_len, proto, group_id);
163 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl);
164 if (__predict_false(ctl == NULL)) {
166 * Can be the case when notification is sent within VNET
167 * which doesn't have any netlink sockets.
175 int io_flags = NL_IOF_UNTRANSLATED;
177 CK_LIST_FOREACH(nlp, &ctl->ctl_pcb_head, nl_next) {
178 if (nl_isset_group_locked(nlp, group_id) && nlp->nl_proto == proto) {
179 if (nlp_last != NULL) {
181 m_copy = m_copym(m, 0, M_COPYALL, M_NOWAIT);
183 nl_send_one(m_copy, nlp_last, num_messages, io_flags);
186 if (nlp_last->nl_socket != NULL)
187 sorwakeup(nlp_last->nl_socket);
188 NLP_UNLOCK(nlp_last);
194 if (nlp_last != NULL)
195 nl_send_one(m, nlp_last, num_messages, io_flags);
203 nl_has_listeners(int netlink_family, uint32_t groups_mask)
205 return (V_nl_ctl != NULL);
212 * app can open multiple netlink sockets.
213 * Start with current pid, if already taken,
214 * try random numbers in 65k..256k+65k space,
215 * avoiding clash with pids.
217 if (nl_port_lookup(curproc->p_pid) == NULL)
218 return (curproc->p_pid);
219 for (int i = 0; i < 16; i++) {
220 uint32_t nl_port = (arc4random() % 65536) + 65536 * 4;
221 if (nl_port_lookup(nl_port) == 0)
223 NL_LOG(LOG_DEBUG3, "tried %u\n", nl_port);
225 return (curproc->p_pid);
229 nl_bind_locked(struct nlpcb *nlp, struct sockaddr_nl *snl)
232 if (nlp->nl_port != snl->nl_pid) {
234 "bind() failed: program pid %d "
235 "is different from provided pid %d",
236 nlp->nl_port, snl->nl_pid);
237 return (EINVAL); // XXX: better error
240 if (snl->nl_pid == 0)
241 snl->nl_pid = nl_find_port();
242 if (nl_port_lookup(snl->nl_pid) != NULL)
244 nlp->nl_port = snl->nl_pid;
245 nlp->nl_bound = true;
246 CK_LIST_INSERT_HEAD(&V_nl_ctl->ctl_port_head, nlp, nl_port_next);
248 for (int i = 0; i < 32; i++) {
249 if (snl->nl_groups & ((uint32_t)1 << i))
250 nl_add_group_locked(nlp, i + 1);
252 nl_del_group_locked(nlp, i + 1);
259 nl_pru_attach(struct socket *so, int proto, struct thread *td)
264 if (__predict_false(netlink_unloading != 0))
265 return (EAFNOSUPPORT);
267 error = nl_verify_proto(proto);
271 bool is_linux = SV_PROC_ABI(td->td_proc) == SV_ABI_LINUX;
272 NL_LOG(LOG_DEBUG2, "socket %p, %sPID %d: attaching socket to %s",
273 so, is_linux ? "(linux) " : "", curproc->p_pid,
274 nl_get_proto_name(proto));
276 /* Create per-VNET state on first socket init */
277 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl);
279 ctl = vnet_nl_ctl_init();
280 KASSERT(V_nl_ctl != NULL, ("nl_attach: vnet_sock_init() failed"));
282 MPASS(sotonlpcb(so) == NULL);
284 nlp = malloc(sizeof(struct nlpcb), M_PCB, M_WAITOK | M_ZERO);
285 error = soreserve(so, nl_sendspace, nl_recvspace);
292 /* Copy so_cred to avoid having socket_var.h in every header */
293 nlp->nl_cred = so->so_cred;
294 nlp->nl_proto = proto;
295 nlp->nl_process_id = curproc->p_pid;
296 nlp->nl_linux = is_linux;
297 nlp->nl_active = true;
298 nlp->nl_unconstrained_vnet = !jailed_without_vnet(so->so_cred);
300 refcount_init(&nlp->nl_refcount, 1);
303 nlp->nl_taskqueue = taskqueue_create("netlink_socket", M_WAITOK,
304 taskqueue_thread_enqueue, &nlp->nl_taskqueue);
305 TASK_INIT(&nlp->nl_task, 0, nl_taskqueue_handler, nlp);
306 taskqueue_start_threads(&nlp->nl_taskqueue, 1, PWAIT,
307 "netlink_socket (PID %u)", nlp->nl_process_id);
310 /* XXX: check ctl is still alive */
311 CK_LIST_INSERT_HEAD(&ctl->ctl_pcb_head, nlp, nl_next);
320 nl_pru_abort(struct socket *so)
322 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid);
323 MPASS(sotonlpcb(so) != NULL);
324 soisdisconnected(so);
328 nl_pru_bind(struct socket *so, struct sockaddr *sa, struct thread *td)
330 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl);
331 struct nlpcb *nlp = sotonlpcb(so);
332 struct sockaddr_nl *snl = (struct sockaddr_nl *)sa;
335 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid);
336 if (snl->nl_len != sizeof(*snl)) {
337 NL_LOG(LOG_DEBUG, "socket %p, wrong sizeof(), ignoring bind()", so);
344 error = nl_bind_locked(nlp, snl);
347 NL_LOG(LOG_DEBUG2, "socket %p, bind() to %u, groups %u, error %d", so,
348 snl->nl_pid, snl->nl_groups, error);
355 nl_assign_port(struct nlpcb *nlp, uint32_t port_id)
357 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl);
358 struct sockaddr_nl snl = {
365 snl.nl_groups = nl_get_groups_compat(nlp);
366 error = nl_bind_locked(nlp, &snl);
370 NL_LOG(LOG_DEBUG3, "socket %p, port assign: %d, error: %d", nlp->nl_socket, port_id, error);
375 * nl_autobind_port binds a unused portid to @nlp
376 * @nlp: pcb data for the netlink socket
377 * @candidate_id: first id to consider
380 nl_autobind_port(struct nlpcb *nlp, uint32_t candidate_id)
382 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl);
383 uint32_t port_id = candidate_id;
386 int error = EADDRINUSE;
388 for (int i = 0; i < 10; i++) {
389 NL_LOG(LOG_DEBUG3, "socket %p, trying to assign port %d", nlp->nl_socket, port_id);
391 exist = nl_port_lookup(port_id) != 0;
394 error = nl_assign_port(nlp, port_id);
395 if (error != EADDRINUSE)
400 NL_LOG(LOG_DEBUG3, "socket %p, autobind to %d, error: %d", nlp->nl_socket, port_id, error);
405 nl_pru_connect(struct socket *so, struct sockaddr *sa, struct thread *td)
407 struct sockaddr_nl *snl = (struct sockaddr_nl *)sa;
410 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid);
411 if (snl->nl_len != sizeof(*snl)) {
412 NL_LOG(LOG_DEBUG, "socket %p, wrong sizeof(), ignoring bind()", so);
417 if (!nlp->nl_bound) {
418 int error = nl_autobind_port(nlp, td->td_proc->p_pid);
420 NL_LOG(LOG_DEBUG, "socket %p, nl_autobind() failed: %d", so, error);
424 /* XXX: Handle socket flags & multicast */
427 NL_LOG(LOG_DEBUG2, "socket %p, connect to %u", so, snl->nl_pid);
433 destroy_nlpcb(struct nlpcb *nlp)
437 NLP_LOCK_DESTROY(nlp);
442 destroy_nlpcb_epoch(epoch_context_t ctx)
446 nlp = __containerof(ctx, struct nlpcb, nl_epoch_ctx);
453 nl_pru_detach(struct socket *so)
455 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl);
456 MPASS(sotonlpcb(so) != NULL);
459 NL_LOG(LOG_DEBUG2, "detaching socket %p, PID %d", so, curproc->p_pid);
462 /* Mark as inactive so no new work can be enqueued */
464 bool was_bound = nlp->nl_bound;
465 nlp->nl_active = false;
468 /* Wait till all scheduled work has been completed */
469 taskqueue_drain_all(nlp->nl_taskqueue);
470 taskqueue_free(nlp->nl_taskqueue);
475 CK_LIST_REMOVE(nlp, nl_port_next);
476 NL_LOG(LOG_DEBUG3, "socket %p, unlinking bound pid %u", so, nlp->nl_port);
478 CK_LIST_REMOVE(nlp, nl_next);
479 nlp->nl_socket = NULL;
485 NL_LOG(LOG_DEBUG3, "socket %p, detached", so);
487 /* XXX: is delayed free needed? */
488 NET_EPOCH_CALL(destroy_nlpcb_epoch, &nlp->nl_epoch_ctx);
492 nl_pru_disconnect(struct socket *so)
494 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid);
495 MPASS(sotonlpcb(so) != NULL);
500 nl_pru_peeraddr(struct socket *so, struct sockaddr **sa)
502 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid);
503 MPASS(sotonlpcb(so) != NULL);
508 nl_pru_shutdown(struct socket *so)
510 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid);
511 MPASS(sotonlpcb(so) != NULL);
517 nl_pru_sockaddr(struct socket *so, struct sockaddr **sa)
519 struct sockaddr_nl *snl;
521 snl = malloc(sizeof(struct sockaddr_nl), M_SONAME, M_WAITOK | M_ZERO);
522 /* TODO: set other fields */
523 snl->nl_len = sizeof(struct sockaddr_nl);
524 snl->nl_family = AF_NETLINK;
525 snl->nl_pid = sotonlpcb(so)->nl_port;
526 *sa = (struct sockaddr *)snl;
531 nl_pru_close(struct socket *so)
533 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid);
534 MPASS(sotonlpcb(so) != NULL);
535 soisdisconnected(so);
539 nl_pru_output(struct mbuf *m, struct socket *so, ...)
542 if (__predict_false(m == NULL ||
543 ((m->m_len < sizeof(struct nlmsghdr)) &&
544 (m = m_pullup(m, sizeof(struct nlmsghdr))) == NULL)))
546 MPASS((m->m_flags & M_PKTHDR) != 0);
548 NL_LOG(LOG_DEBUG3, "sending message to kernel async processing");
549 nl_receive_async(m, so);
555 nl_pru_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *sa,
556 struct mbuf *control, struct thread *td)
558 NL_LOG(LOG_DEBUG2, "sending message to kernel");
560 if (__predict_false(control != NULL)) {
561 if (control->m_len) {
568 return (nl_pru_output(m, so));
572 nl_pru_rcvd(struct socket *so, int flags)
574 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid);
575 MPASS(sotonlpcb(so) != NULL);
577 nl_on_transmit(sotonlpcb(so));
583 nl_getoptflag(int sopt_name)
586 case NETLINK_CAP_ACK:
587 return (NLF_CAP_ACK);
588 case NETLINK_EXT_ACK:
589 return (NLF_EXT_ACK);
590 case NETLINK_GET_STRICT_CHK:
598 nl_ctloutput(struct socket *so, struct sockopt *sopt)
600 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl);
601 struct nlpcb *nlp = sotonlpcb(so);
603 int optval, error = 0;
606 NL_LOG(LOG_DEBUG2, "%ssockopt(%p, %d)", (sopt->sopt_dir) ? "set" : "get",
607 so, sopt->sopt_name);
609 switch (sopt->sopt_dir) {
611 switch (sopt->sopt_name) {
612 case NETLINK_ADD_MEMBERSHIP:
613 case NETLINK_DROP_MEMBERSHIP:
614 error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval));
617 if (optval <= 0 || optval >= NLP_MAX_GROUPS) {
621 NL_LOG(LOG_DEBUG2, "ADD/DEL group %d", (uint32_t)optval);
624 if (sopt->sopt_name == NETLINK_ADD_MEMBERSHIP)
625 nl_add_group_locked(nlp, optval);
627 nl_del_group_locked(nlp, optval);
630 case NETLINK_CAP_ACK:
631 case NETLINK_EXT_ACK:
632 case NETLINK_GET_STRICT_CHK:
633 error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval));
637 flag = nl_getoptflag(sopt->sopt_name);
641 nlp->nl_flags |= flag;
643 nlp->nl_flags &= ~flag;
651 switch (sopt->sopt_name) {
652 case NETLINK_LIST_MEMBERSHIPS:
654 optval = nl_get_groups_compat(nlp);
656 error = sooptcopyout(sopt, &optval, sizeof(optval));
658 case NETLINK_CAP_ACK:
659 case NETLINK_EXT_ACK:
660 case NETLINK_GET_STRICT_CHK:
662 optval = (nlp->nl_flags & nl_getoptflag(sopt->sopt_name)) != 0;
664 error = sooptcopyout(sopt, &optval, sizeof(optval));
678 sysctl_handle_nl_maxsockbuf(SYSCTL_HANDLER_ARGS)
681 u_long tmp_maxsockbuf = nl_maxsockbuf;
683 error = sysctl_handle_long(oidp, &tmp_maxsockbuf, arg2, req);
684 if (error || !req->newptr)
686 if (tmp_maxsockbuf < MSIZE + MCLBYTES)
688 nl_maxsockbuf = tmp_maxsockbuf;
694 nl_setsbopt(struct socket *so, struct sockopt *sopt)
699 if (sopt->sopt_name != SO_RCVBUF)
700 return (sbsetopt(so, sopt));
702 /* Allow to override max buffer size in certain conditions */
704 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
707 NL_LOG(LOG_DEBUG2, "socket %p, PID %d, SO_RCVBUF=%d", so, curproc->p_pid, optval);
708 if (optval > sb_max_adj) {
709 if (priv_check(curthread, PRIV_NET_ROUTE) != 0)
713 SOCK_RECVBUF_LOCK(so);
714 result = sbreserve_locked_limit(so, SO_RCV, optval, nl_maxsockbuf, curthread);
715 SOCK_RECVBUF_UNLOCK(so);
717 return (result ? 0 : ENOBUFS);
720 #define NETLINK_PROTOSW \
721 .pr_flags = PR_ATOMIC | PR_ADDR | PR_WANTRCVD, \
722 .pr_ctloutput = nl_ctloutput, \
723 .pr_setsbopt = nl_setsbopt, \
724 .pr_abort = nl_pru_abort, \
725 .pr_attach = nl_pru_attach, \
726 .pr_bind = nl_pru_bind, \
727 .pr_connect = nl_pru_connect, \
728 .pr_detach = nl_pru_detach, \
729 .pr_disconnect = nl_pru_disconnect, \
730 .pr_peeraddr = nl_pru_peeraddr, \
731 .pr_send = nl_pru_send, \
732 .pr_rcvd = nl_pru_rcvd, \
733 .pr_shutdown = nl_pru_shutdown, \
734 .pr_sockaddr = nl_pru_sockaddr, \
735 .pr_close = nl_pru_close
737 static struct protosw netlink_raw_sw = {
742 static struct protosw netlink_dgram_sw = {
743 .pr_type = SOCK_DGRAM,
747 static struct domain netlinkdomain = {
748 .dom_family = PF_NETLINK,
749 .dom_name = "netlink",
750 .dom_flags = DOMF_UNLOADABLE,
752 .dom_protosw = { &netlink_raw_sw, &netlink_dgram_sw },